Japanese art house favorite Hirokazu Kore-eda is a somewhat unlikely figure to probe the sci-fi implications of generative artificial intelligence. The 63-year-old auteur, winner of the Cannes Palme d’Or in 2018 for Shoplifters, has made his indelible imprint on world cinema with delicate family drama, suffused with wry humor and wrenching humanism, far more so than futurism. But for his 17th feature, Sheep in a Box, Kore-eda has set his story in a speculative world just over the horizon — where packages are delivered by drone, all cars are electric, and generative AI has reached into the most intimate domains of human experience.

The film stars Haruka Ayase and Daigo Yamamoto as a young couple mourning the recent death of their son. They live in a leafy exurb of Tokyo in a sleek, sunlit home designed by the wife, an architect, and built by the husband, a carpenter with a soulful reverence for high-quality wood. But the weight of their grief is just as palpable as the utopian patina of their surroundings. A dubious promise of relief arrives by way of a new AI-powered robotics company — one that specializes in vividly realistic android re-creations of lost loved ones. Tentatively, the bereaved parents soon welcome a little humanoid (played by newcomer Kuwaki Rimu) into their home who is indistinguishable from their beloved late son, Kakeru — except for a power button on the back of his neck and a nightly need to sit on his charging station.

In other hands, the stage would be set for a Black Mirror-like dystopian chiller, but Kore-eda, unsurprisingly, takes his premise into less obvious, more nuanced emotional terrain. You could say one of world cinema’s great humanists extends the grace of his embrace into the realm of the post-human.

The Hollywood Reporter connected with Kore-eda in Tokyo ahead of Cannes to discuss how Sheep in a Box came about.

Tell me about the creative origins of this project. Where did this premise come from?

In general, I’ve gotten interested in generative AI. But the more specific spark for this story came to me in March of 2024, when I happened upon an article about a Chinese startup [called Super Brain] that uses AI to resurrect people who have passed away. That got me even more interested, and I wrote up a little treatment. Later that fall, I was on a trip to Beijing for other purposes and I arranged an opportunity to speak to the company’s founder, Zhang Zewei. He gave me a demo of how their service works. Basically, they use a bunch of data about the deceased person, including audiovisual data — photos and videos — they create an AI likeness of the deceased person that their loved ones can interact with. What I found fascinating is that the technology makes it possible to have new conversations with them — not just covering topics you’ve discussed with them before. I found it all precarious, but I could also see how this is something that will inevitably spread. So I decided to expand my treatment and explore these issues more deeply.

I was struck by how Sheep in a Box is a pretty optimistic take on AI overall. It’s not particularly dystopian. How did you come to that point of view?

I can’t say I’m especially well-read in the sci-fi space, but when you consider Isaac Asimov’s famous Three Laws of Robotics [which can be paraphrased as: A robot may not harm a human being; a robot must obey human orders, unless those orders conflict with the first law; and a robot must protect its own existence, unless doing so conflicts with the first or second law] they are very anthropocentric. It’s all based on the idea that humans will always be the center of this world. And I’ve always felt a little discomfort with that approach. I believe that as AI and androids evolve, they are going to transcend humanity, and at that point, humanity is not going to be something they even really care about. They will want to connect with something bigger. That’s what I see as the most realistic outcome, anyway.

So, one of the first thoughts was that I wanted my story to land in this notion of androids choosing to no longer exist among the humans. Then I began thinking of how children always outgrow and transcend their parents — you know, the idea of them leaving the nest, and eventually living lives that the parents sometimes struggle to keep up with. I decided to layer these two stories, and that’s essentially how it came about.

How did you approach the world-building aspect of the project — deciding just how futuristic the film’s pseudo sci-fi setting would be?

One of the big themes of the film is the idea of the box, so a key breakthrough for the production was finding the modernist house that we shot as the couple’s home. It’s composed of overlapping boxes, and if you take a bird’s-eye view of the whole structure, it’s essentially a box with a square garden courtyard in the middle. Finding that house and drawing various ideas from it was huge for the production. Not just in terms of scene details or art direction, but because the sets themselves were built to align with the layout of the actual house, and I revised some things in the script after taking inspiration from the architecture.

The house itself is in Kamakura, which is a bit away from Tokyo and very different from the lived-in, downtown feel that I portrayed in Shoplifters. Kamakura is a bit high-end, and it has a feeling of being more open and connected to nature. In that sense, both the town and the house are quite sophisticated.

The real home was built by a couple who live there with their kids. The wife is an architect and the husband works for a construction company — just like the characters in our story. We rented the house as it is. It was foundational to the whole film.

We also witness the mother’s creative process as an architect throughout the story. She uses AI regularly, but she also continues to work in ways that are resolutely analog. She makes meticulous paper models of her buildings, and there’s a moment when she snaps at the android little boy when he interferes in her process and tries to just tell her the answer to her creative experimentation. She tells him, “Don’t take that part away.” I sensed some potential notes of instruction on your part here — about how our relationship with AI might evolve, or should evolve, if we’re lucky and thoughtful.

Something that I often think about these days is process. In Japanese, the word we might use to describe this kind of activity is muda, which translates to waste, futility or effort that doesn’t really yield any direct value. But I feel that the time we spend in that state is what makes us human. So, with that scene you mentioned, I wanted to give some hints of that. But if you look at the whole story from a broader perspective, the couple go through an evolution of processing their grief, their relationship, and how they relate to and feel about this android version of their son, and it’s precisely that process — the search, and the trial and error of it — that makes the film a humanistic portrayal. AI offers the promise of basically just being able to present you with the answer. In many contexts, that will certainly save time — it eliminates the muda — but it just doesn’t feel as good, ultimately. There’s no merit in it. It’s like being given the answer without playing the game.

Lastly, I have to ask … Does Hirokazu Kore-eda use AI and what’s your relationship with it like — in your work or daily life?

Personally, no, I don’t use it at all. Not at all. But while I was making this film, I thought, “Why not give it a go?” So I asked a member of my crew to have ChatGPT read my script and evaluate it. We explained our goals and asked it, “What are some ideas you may have to make this script better?” I was hoping to have a productive back and forth. And it was OK. It was interesting. I could see how it has become something that feels fun to talk to — but it didn’t give me any unexpected answers. Maybe someday it will evolve to the point where it has the capability to give you something really surprising and compelling, but that wasn’t my experience.